You are on page 1of 9

Proceedings 7th Modelica Conference, Como, Italy, Sep.

20-22, 2009

Modeling of Rotary Kilns


and Application to Limestone Calcination
Uwe Kssel1 Dirk Abel1 Matthias Schumacher2 Martin Weng2
1 RWTH Aachen University, Institute of Automatic Control
Steinbachstrae 54A, D-52074 Aachen
2 Aixprocess, Process and Fluid Engineering
Alfonsstrasse 44, D-52070 Aachen

Abstract

become available to close the control loop. To be applicable to an expert system, the very complex model
This paper presents the one dimensional modeling of of the process has to be capable of real-time operation
rotary kilns used for energy intensive production pro- anyway and thus, several assumptions and simplificacesses. Raw material is fed into an inclined rotating tions have to be done.
kiln and heated by counter current gas flow. Chemical
As a first step to the automatic control of continureactions take place in the bed of raw material as well ously operating rotary kilns, a detailed model of this
as in the gas phase. Heat and mass transfer between process is developed. The model abstracts from rethe bed and the gas phase are implemented. Also the ality by assuming one dimensional (1-D) counter curheat transfer to the environment is taken into account. rent flow of the raw material phase (bed phase) and the
As a benchmark, the process of limestone calcination gas phase. The bed and gas phase are surrounded by
is chosen. Results are compared with computational a combination of isolating refractory and steel shells.
fluid dynamic simulations.
Two reactions are implemented, one for the production
Keywords: CaCO3 , calcination, CFD comparison, of heat and one for the calcination itself. Methane is
kilns, limestone, rotary kilns, simulation
oxidized in the gas phase in order to supply the neces-

Introduction

Unhydrated lime is used as a raw material for many


products in chemical industry. The limestone calcination, as an energy intensive production process for
unhydrated lime, is often performed in continuously
operating rotary kilns. Until today, the process is manually operated and despite existing approaches the use
of automatic control is very uncommon. Due to the hot
and dusty atmosphere inside the drum, thermodynamic
states used as controlled variables by an expert system
are hardly measurable in a reliable way. In heat driven
chemical production processes such as limestone production measured data like a temperature as one thermodynamic variable of the process (often most easily
to measure) is not sufficient to estimate the chemical
rate of degradation of CaCO3 . To overcome this problem dynamic physical and chemical models can be applied to predict the behavior of dependent measures.
By comparing the results of the modeling to current
plant measurements the states of independent variables

The Modelica Association, 2009

sary energy to the process and limestone is calcinated


in the bed phase consuming energy due to an endothermic degradation process. Heat and mass transfer take
place between the counter current flows and the environment. The paper is structured as follows: In chapter
2 the abstraction to 1-D counter current flow is shown.
Additionally, the mechanisms of reactions and heat
transfer are described. In chapter 3 the computational
results with Dymola are analyzed in detail. For validation, the results are compared to computational fluid
dynamics (CFD) simulations in chapter 4. Chapter 5
concludes the paper with an outlook of future investigation.

Modeling of rotary kilns

The rotary kiln is modeled using a 1-D approach. The


flows of gas and bed phase are counter current. The
rotary kiln itself consists of an isolation and a steel
shell. This abstraction is depicted in Figure 1.

814

DOI: 10.3384/ecp09430084

Proceedings 7th Modelica Conference, Como, Italy, Sep. 20-22, 2009

Figure 1: Rotary kiln with 1-D balance element

2.1

Modeling of the gas phase

The gas phase model is based on the DynamicPipe


model of the Modelica fluid library [1]. It is extended
by dynamically changing cross sectional areas due to
bed heights. Replaceable models of radiative and convectional heat transfer are integrated and will be discussed in section 2.6. In addition, the gas phase is
equipped with a replaceable chemical reaction model
in order to cover the oxidation of mixed gaseous hydrocarbons (i.e. in this case Methane only), details are
covered in section 2.5.
The balancing equations with a dynamic momentum balance are formulated in [2]. Using a finite volume approximation and setting the momentum balance to be static, the equations (1)-(3), corresponding
to a single slice used in the DynamicPipe model, are
derived. To avoid obscurity, additional equations covering multi-component media are not given in this formulation.
dmi
dt

= m i+ 1 + m i 1 +
2

hydrocarbons are added to the RHS of the mass balance in equation (1). Including the heat of formation
of all components implicitly adds the contribution of
chemical reactions to the energy balance while only
considering mass balance changes. For the longitudinal gas flow and the heat transfer standard connectors
are used. For the gaseous exchange between gas and
bed signal oriented connectors for each direction are
designed using the same exchange medium (O2 , CO,
CO2 , H2 O).
All cross sectional areas, and hence corresponding
volumes, in these equations are dynamically changed
due to changes in bed height. Therefore, the information of the beds cross sectional area is transmitted via
input output relation to the gas phase. Finally, data
on gas phase properties are provided for other components by an output connector.

2.2

Modeling of the bed phase

The bed phase is modeled by mass and energy balances.


dmi
dt

m i+ 1 + m i 1 +
2

(4)

with m i+ 1 = iVi+ 1
2

dUi
(5)
= H i+ 1 + H i 1 +
2
2
dt
Heat and mass transfer is included similarly to gas
phase modelling via source terms to the RHS of the
balance equations. The transport of material through
the rotary kiln is specified by equation (6), which is
commonly known as Kramers equation [5].

(1)

)3
(
)(
2 2
4
tan
dh
h
h
i
i
3

2
2
2 i2
0 =

(2) Vi+ 12 = 3 R sin dx cot


R R
Ai+ 1 pi+ 1 Ai 1 pi 1
2
2
(6)
) 2
( 2
The volume flow rate is a function of the rotation
+Ai pi+ 1 pi 1
2
2
frequency , the inner radius of the pipe R, the inclinaFFFV Ai i g z
tion of the kiln , the materials angle of repose , the
dUi
dh
= H i+ 1 + H i 1
(3) height of material h and the corresponding gradient dx .
2
2
dt
This differential equation in terms of the height is sim(
)
+vi Ai pi 1 pi+ 1 +
plified by rearranging and approximating the height
2
2
and the corresponding gradient with the equations (7)
for slice i = 1, . . . , n
to (9).
Source terms for heat, namely radiational and convectional contributions, are added to the right hand
hi = R (R cos (i /2))
(7)
side (RHS) of the energy balance in equation (3). Mass
dhi
di
= (R/2) sin (i /2)
(8)
transfer between bed and gas phase due to degradation
dx
dx
of CaCO3 (releases CO2 ) and changes in the compodi
(i+1 i ) / (L/n)
(9)
sition of the gas phase due to the oxidation of gaseous
dx
m 2i+ 1

The Modelica Association, 2009

m 2i 1

815

Proceedings 7th Modelica Conference, Como, Italy, Sep. 20-22, 2009

The mid point angle of the bed phase is denoted by


The error between the approximation and original
, the length of the rotary kiln by L and n is the number data pairs is given in Figure 2.
of slices in the rotary kiln. All geometric details for
gas and bed phase are calculated using the mid point
angle. The calculation of i = f (Ai ) is presented in
section 2.4.

2.3

Modeling of the shells

The isolation and steel shells are modeled by an energy balance for the corresponding material volume
resulting from the 1-D balancing elements. Heat transfer, and hence the temperature profile, is realized using
Fouriers relation for heat conduction in solid material.
In principle, the shells are structured as grids consisting of heat capacitors and thermal conductors likewise
modeled in the modelica standard package. Heat transfer is possible in two directions, namely longitudinally
Figure 2: Approximation error for = f (P)
and transversely to the flow direction in the rotary kiln.
Geometric and material properties are parameters of
The adapted relationship = f (P) is implemented
the model. Standard heat connectors are used.
as a function and integrated in both the gas and bed
phase in order to calculate the mid point angle and
2.4 Mid point angle approximation
subsequently all geometric characteristics of the rotary
kiln.
All geometric properties for gas and bed phase can
easily be calculated using the mid point angle of the
bed phase. Nevertheless, the angle has to be calculated 2.5 Mechanism of chemical reactions
from the cross sectional area of the bed phase, which
is a direct result of the mass balance in every bed slice The replaceable chemical reaction model is designed
as a reactant limited elementary reaction using an Arof the rotary kiln. This is shown in equation (10).
rhenius type approach for the reaction kinetics. The
Ai = (mi /i ) / (L/n)
(10) reaction kinetics approach is shown with the equation
(14).
The relation between the cross sectional area Ai of
a single slice and the mid point angle i is implicitly
Ea
k = B T n e RT
(14)
given with the equations (11)-(12).
The preexponential coefficient B, the activation en(11) ergy Ea and n are free parameters which will be used
P = 2Ai /R2
(12) to fit the model to CFD simulation data. Generally
speaking, this fit is necessary since, on the one hand,
In order to avoid additional nonlinear equations, the no reliable chemical reaction data is available for such
relationship given with equation (11) is numerically a set up of coupled ideal reactors and, on the other
solved for in a range of P between 1.25 105 and hand, the model is not able to cope with local condi2. The corresponding pairs are then used for a least tions due to low resolution and the lack of adequate
square fit in order to find a good approximation for models. For example, the oxidation of mixed gaseous
= f (P). The function = f (P) is given with equa- hydrocarbons (i.e. in this case CH4 only) is mass transtion (13).
fer limited and therefore, an ideal mixed reactor needs
to be adapted using the averaged chemical reaction kinetics.
1
= a0 log(P) + a1 (1/P) + a2 P 2
The rate of degradation (massflow) for each com+a3 P + a4 P2 + a5 P3 + a6 P4
(13) ponent is calculated using the relationship given with
+a7 P5 + a8 P6 + a9
equation (15).
0 = P + (i sin i )

The Modelica Association, 2009

816

Proceedings 7th Modelica Conference, Como, Italy, Sep. 20-22, 2009

Within small deviations in the range of Pr,


0.5 < Pr < 1.5, the simplified equation (21) can be
(15)
used.

k=1

m j = kV j M j Ckk
l

In this relationship, the velocity of reaction k is coupled with the volume of the reaction element V , the
molar mass M and the stoichiometric cofficient for
each component (index j over all components). The
approach is reactant limited assuming an elementary
reaction order. Hence, the molar density C to the
power of the stoichiometric coefficient sums up over
all reactants k of a reaction.
For the bed phase, the degradation of CaCO3 is implemented. The chemical reaction is given with equation (16).
CaCO3 = CaO +CO2

(16)

Energy is chemically provided by the combustion of


CH4 , equation (17) holds.
CH4 + 2O2 = CO2 + 2H2 O

(17)

In principle, various reactions are possible, as long


as the defined medium covers all components and the
stoichiometric coefficients are defined.

2.6

Mechanisms of heat transfer

For the heat transfer between the gas and bed phase
as well as the isolation shell convectional and radiational mechanisms are implemented. For convectional
heat transfer between the gas and its corresponding exchange partner, the heat transfer coefficient is dependent on the Reynolds number Re, the Prandtl number Pr and the Nusselt number Nu. The general convectional heat transfer equation in terms of is shown
with equation (18).
Q conv 12 = 12 A12 (T1 T2 )

[
( ) 23 ]
( 0.8
) 0.4
Dh
Nu = 0.0214 Re 100 Pr
1+
L
(21)
All the models calculate in a similar range from
8 10 W/m2 K.
Convectional heat transfer between the bed phase
and the isolation is also realized with the heat transfer coefficient and equation (18). is calculated
from various process parameters using two different
approaches which are described in detail in [8]. Typical values are said to be 50 200 W/m2 K
Also, models with constant are possible to
choose. The heat transfer models are designed to be
replaceable in order to enable the choice between different heat transfer mechanisms as described above.
Radiation between the bed, the gas and the isolation
is modeled using equation (22).
(
)
Q rad 12 = 12 A12 T14 T24

(22)

The referenced area for radiational heat transfer between transfer object 1 and 2 is given with A12 . The
emissivity coefficient 12 relates the visibilty between
area A12 and A21 and the emissivity of corresponding
objects. There are three emissivity coefficients to be
calculated, namely wall bed (wb), wall gas (wg)
and bed gas (bg). The formulas for calculating
these coefficients are given with equation (23) to (26).

(18)

Various heat transfer models for the gas phase solid


interaction are implemented in the Modelica fluid
library [1]. Additionally, two relationships for Nusselt
numbers are given in [10]. The equations are shown
with (19) and (20).

w b (1 g )
U
w g (1 + (1 g ) (1 b ))
=
U
b g (1 + (1 g ) (1 w ))
=
U
= 1 (1 g ) (1 w )

wb =

(23)

wg

(24)

bg
U

(1 (1 (1 b ) (1 g )))

(25)
(26)

The variable is defined as the ratio between the


) 32 ]
free bed area (surface) and the free isolation area. De /8 (Re 1000) Pr
Dh

Nu =
(19) tails about the derivation can be found in [4]. The
(
) 1+
L
1 + 12.7 /8 Pr2/3
values of emissivity are almost constant over length x
= (1.82 log10 (Re) 1.64)2
(20) and temperature T with wb = 0.184, wg = 0.695 and
2300 < Re < 106 , Dh /L < 1
bg = 0.615.
[

The Modelica Association, 2009

817

Proceedings 7th Modelica Conference, Como, Italy, Sep. 20-22, 2009

Computational results

The model is designed and numerically integrated


within the Dymola modeling environment. The set up
is depicted in the Figure 3.

Figure 4: Degradation of CaCO3 through calcination


increase of temperature and the endothermic character
of the calcination reaction. The increase of temperature for the main components of the rotary kiln (steel,
isolation, gas and bed phase) is depicted in Figure 5

Figure 3: Dymola model of a rotary kiln


The black and the grey objects denote the steel and
isolation shell, respectively, with different geometric
and material parameters. Inside the pipe, there is
counter current flow of the gas phase (blue element)
and the bed phase (limestone material, grey element).
Different mechansims for heat and mass transfer as
well as chemical reactions are chosen as described
earlier in this paper. Further components, e.g. crust
of semi liquefied limestone, can be added but are not
modeled in this contribution.
The calcination process (degradation of CaCO3 ) is
shown in Figure 4.
The system is initialized at 273.15 K and atmospheric pressure (time [s]: 0-3000). In a next step, the
system is brought to the operating point by increasing
temperature of the incoming mass flows of the bed and
the gas phase (time [s]: 3000-13000). This procedure
is followed by the intial increase of the combustion
gas massflow (time [s]: 13000-250000, massflow step
[kg/s]: 0.1-1.2692). It is clear to see that the rate of
CaCO3 degradation intensifies as the temperature increases. A second step to the combustion gas massflow
is applied after 300000 seconds (massflow step [kg/s]:
1.2692-2.5392). Again, the degradation of CaCO3 increases. The increase of degradation, while applying
these steps to the combustion gas flow, is due to the

The Modelica Association, 2009

Figure 5: Temperature of steel, isolation, bed and gas


phase
While the gas phase (red) responds extremely fast
to the step in combustion gas massflow, the bed phase
(yellow) is slower. The isolation (turquoise) and steel
(blue) shell are even slower in their step response. The
different time constants of the process can be recognized by an exemplary inspection of Figure 6 to 7.
It is easy to observe the small time constant of the
gas response lying in the range of 10-200 seconds.
For the isolation shell, the time constant of reponse
lies in the range of 20000-30000 seconds. This explains the stiffness of the system and the necessity of
stiff numerical solvers for simulations. Furthermore,
these time constants correspond to the known time
constants from process industry, which implies that

818

Proceedings 7th Modelica Conference, Como, Italy, Sep. 20-22, 2009

stone particles and the coupling of gas and solid phase.


Since this work mainly focuses on the 1-D model of
the limestone calcination process, only a short summary of the models used in the CFD simulations will
be given here.

4.1

Introduction of the CFD rotary kiln


model

The three dimensional computational domain of the


freeboard in the kiln is bounded by the refractory and
the surface of the solid bed. The methane burner extends into the kiln on one end of the drum (Figure 8).
Figure 6: Temperature response of gas phase due to
combustion gas step

Figure 8: 3-D computational domain of the limestone


rotary kiln
Figure 7: Temperature response of isolation due to
combustion gas step
the dynamical behaviour of the model reliably represents the calcination process. Although the steady
state values of the operating point also correspond to
the known values from process industry, this modeling
approach is justified even more by comparing its values to CFD simulation results in the following chapter.

Comparison to CFD simulations

Computational fluid dynamics (CFD) is a widely accepted tool for the detailed description of gaseous
combustion phenomena occurring within a rotary kiln
[6],[3]. Unfortunately, it does neither allow for the
modeling of transport of the solid bed nor for the theoretical description of the chemical reactions therein. In
the current work, different sub-models have been integrated in the commercial CFD code Fluent in order to
model the dynamics of the granular flow of the lime-

The Modelica Association, 2009

Steady state conservation equations of the compressible gas flow are solved considering the realizable k--model for turbulent effects in the gas phase.
Transport of four separate species (O2 , CH4 , CO2 and
H2 O) is calculated explicitly in the gas phase, while
the fifth one (N2 ) sums to unity. Combustion rates
of methane oxidization are assumed to be limited by
the mixing of turbulent eddies and thus, the eddydissipation model is used to predict reaction rates of
the volumetric reactions in the gas phase. Energy
transport phenomena include conduction, convection
and radiation (P-1 model), while the latter plays the
major role in rotary kiln processes. Absorption of the
gas mixture is mainly affected by the product constituents of the gas (CO2 and H2 O). Therefore the
wsggm-cell-based model is applied to calculate the local absorption coefficient of the gas mixture. A significant amount of energy discharges into the environment. Conduction driven heat transfer through the refractory is being calculated explicitly by applying a
computational grid to this region also. On the outer

819

Proceedings 7th Modelica Conference, Como, Italy, Sep. 20-22, 2009

surface of the kiln two heat transfer mechanisms are


considered: convection and radiation.
Thermal coupling between gas and solid phase is realized by setting a temperature profile on the bed surface. For this purpose mass, species and energy conservation equations within the particulate solid bed are
solved by implementing a mathematical submodel for
the solid bed. Since it has recently been proven, that
axial mixing can be neglected in industrial rotary kiln
processes, transport of the solid bed can be simplified
by assuming a plug flow [9]. Additionally the bed is
assumed to be well mixed locally in any given cross
section.
Limestone calcination can be modeled as a shrinking core process with surface reaction control according to equation (27) [7].

(
)
dmCaCO3
EA
= k0 exp
4 r2p Np MCaCO3
dt
R Tp
(27)
Herein r p and Np represent the mean particle radius and the total number of limestone particles, respectively. CO2 released by the calcination reaction is
assumed to be transported instantaneously to the hot
flow by neglecting any transport resistance in the solid
bed. It is released in the computational space of the
hot gas flow by adding mass and species sources in the
cells adjacent to the bed surface.
In Figure 9 the contour plot of the local gas temperature on the center plane is shown as an exemplary
result of the CFD simulations.

Figure 9: Contour plot of the local gas temperature on


the center plane in [K]
In order to justify the 1-D abstraction of the process
modeled within the Dymola environment, the computational results are compared to the highly resolute results of the CFD simulations. For this purpose, volume
V, j of the CFD simulation data
based average values
are calculated for a discrete number of slices in the gas
phase. Due to the strongly anisotropic flow field in the
vicinity of the burner mass and velocity weighted average has to be applied for the potential variables. This
is shown in equation (28).

The Modelica Association, 2009

Av
V, j = Ncells i i i i,ax

Ncells i Ai vi,ax

(28)

In equation (28), Vi as the volume of each single


cell inside the discrete slice and vi,ax as the local axial
velocity of the gas mixture are used to calculate the
volume based average of slice j for each scalar .
Since plug flow has been assumed in the clinker bed,
equation (28) can be simplified to the volume weighted
average for all potential variables in the solid phase as
given in equation (29).
Ncells i Vi
s
V,

j=
Ncells Vi

(29)

In contrast, flow variables such as reaction rates


need to be integrated within each single control volume in the gas and in the solid domain, respectively.
For these data equation (30) is valid.
V, j =

(30)

Ncells

4.2

Comparison

Geometrical parameters and boundary conditions are


chosen to be equal in both simulation environments.
In addition, the free chemical reaction parameters
(B, n, Ea ) for adapting the model are fitted to CFD simulation data by applying a least square fit using equation (15). Therefore, the CFD simulation data are concentrated in larger 1-D cells by applying the above described weighting functions. After model adaption via
least square fit, a Dymola simulation to steady state is
performed. The simulated data from the Dymola simulation environment are then compared to the weighted,
concentrated CFD simulation data. The essential values of the process are compared. This includes temperature profile of gas and bed phase, molar densities
of O2 and CH4 in the gas phase as well as molar density of CaCO3 in the bed phase (controlled variable for
future model predictive control).
Figure 10 shows the temperature profiles of the gas
phase for Dymola and CFD simulation.
In the next figure (Figure 11), the consumption of
O2 and CH4 is depicted.
The figures show a similar behaviour of the gas
phase in both simulation environments. For the error
e holds e < 2%. On first inspection, this seems to be
a good result. Nevertheless, the more relevant data for
intended controlling of the rotary kiln process is the
temperature and the concentration of components in

820

Proceedings 7th Modelica Conference, Como, Italy, Sep. 20-22, 2009

Figure 10: Comparison of gas phase temperature

Figure 12: Comparison of bed phase temperature

Figure 11: Comparison of gas phase molar density of Figure 13: Comparison of bed phase molar density of
O2 and CH4
CaCO3
the bed phase. Figure 12 shows the temperature of the
bed phase.
Despite the different scaling of the graph, for the error e holds e < 3%. The next figure (Figure 13) shows
the degradation of CaCO3 in the bed phase using the
molar density as depicted variable. The maximum error for the molar density of CaCO3 is e = 6.6%. Although this value is slightly higher than the previous
errors, the 1-D modeling approach is successfully applied. Various alternative values of the process were
compared, yielding to the same outcome. The main
advantage of the 1-D modeling approach is the possiblity of faster simulations while deviations are kept
small enough in the scope of robustness in terms of observer based model predictive control. While the simulation until convergence in the CFD environment takes
up to several days, the simulation of the Dymola model
takes seconds to minutes depending on the change of
input variables. Another enormous advantage of the 1-

The Modelica Association, 2009

D modeling approach is the use of the 1st principles for


physical abstraction. Since the identified data based
models are only valid in the range of measured data,
the nonlinear 1-D model of the rotary kiln will have a
wider range of validation. Furthermore, the simulated
data sets are always open to physical interpretation and
hence, are easier to check for plausibility.

Outlook

In a first step, the complex counter current flow process of rotary kilns is modeled and applied to the limestone degradation. First results show reasonable behaviour of the process physics. The comparison to
CFD simulation data confirms these results. The essential process parameters are compared and the errors
are small enough to allow for application in observer
based model predictive control application. In addi-

821

Proceedings 7th Modelica Conference, Como, Italy, Sep. 20-22, 2009

tion, the modeling approach has a lower computational [8] A. Queck. Untersuchung des gas- und wandseitigen Wrmetransportes in die Schttung von
burden and results in linearizations which are capable
Drehrohrfen. PhD thesis, Otto-von-Guerickeof real-time application in an expert system.
Universitt Magdeburg, 2002.
For the future, details in both models will be refined
in order to get even better results. Furthermore, the
model is going to be enhanced with a particle burner [9] R.G. Sheritt, J. Chaouki, A. Mehrotra, and L. Behie. Axial dispersion in the three-dimensional
and applied to the chemically more complex cement
mixing of particles in a rotating drum reactor.
production process. The task of automatic least square
Chemical Engineering Science, 58(2):401415,
fit for the free chemical parameters of the model will
2003.
be enhanced in order to cover the more complex cement production process. A detailed report on both [10] VDI-Gesellschaft
Verfahrenstechnik
und
topics will soon be published. Alongside with the enChemieingeniuerwesen,
editor.
VDIhanced model, linearizations of the nonlinear model in
Wrmeatlas. Springer, 8th edition, 1997.
desired operating points will be derived. Using these
linearizations, an observer for immeasurable process
values will be established. Furthermore, a model pre- 6 Acknowledgement
dictive control will be designed in order to hold a deThe authors gratefully thank the Federal Ministry
sired operating point of the plant.
of Education and Research for funding BMBF 3257
MoProOpt.
Furthermore, we highly appreciate
References
the supervision by the Project Management Agency
Forschungszentrum Karlsruhe (PTKA).
[1] F. Casella, H. Tummescheit, and M. Otter.
Modelica fluid library
www.modelica.org/libraries/
modelica_fluid/releases/1.0, 2009.
[2] H. Elmqvist, H. Tummescheit, and M. Otter.
Object-oriented modeling of thermo-fluid systems. Modelica Association, November 2003.
[3] M. Georgallis, P. Nowak, M. Salcudean, and
I.S. Gartshore. Modelling the rotary lime kiln.
Canadian Journal of Chemical Engineering,
83(2):212223, 2005.
[4] R. Jeschar, R. Alt, and E. Specht. Grundlagen
der Wrmebertragung. Viola-Jescher Verlag,
1990.
[5] H. Kramers and P. Croockewit. The passage
of granular solids through inclined rotary kilns.
Chemical Engineering Science, 1(6):259265,
1952.
[6] F. Marias, H. Roustan, and A. Pichat. Modelling
of a rotary kiln for the pyrolysis of aluminium
waste. Chemical Engineering Science, 60:4609
4622, 2005.
[7] K.S. Mujumdar, K.V. Ganesh, S.B. Kulkarni,
and V.V. Ranade. Rotary cement kiln simulator
(rocks): Integrated modeling of pre-heater, calciner, kiln and clinker cooler. Chemical Engineering Science, 62:25902607, 2007.

The Modelica Association, 2009

822